Low-income households struggle to accumulate emergency savings, which increases economic vulnerability in the face of unexpected events like expensive car repairs. This vulnerability may be even greater among persistently low-income households, which might benefit most from building emergency savings using tax refunds. This study examined the effects of randomly assigned behavioral interventions that incorporated a choice architecture manipulation and savings-related messages aimed at encouraging refund saving and delivered through a free, online tax-filing software program. The study sample comprised 4,536 tax filers, including 1,235 with persistent low incomes. Using administrative tax data and data from a two-wave household financial survey, regression-adjusted treatment impacts were estimated using intent-to-treat analysis to examine whether filers had any of their tax refunds still in savings and how much of their refund they still had saved six months after filing their taxes. Results indicated directional but nonstatistically significant increases in these savings outcomes across three treatment groups for the full sample, yet statistically significant treatment effects among the persistently low-income subsample, effects that were moderated by the prefiling absence of emergency resources. These results suggested that tax-time savings interventions are most effective among low-income households with the greatest needs for emergency savings.